Industry
Generative AI and the Conflict Within the Fanfiction Community
An exploration of the tension between human authors and generative AI in fanfiction communities and the futility of AI detection tools.
Latest
277 stories in the archive
Models
Mistral's Leanstral 1.5: Scaling Mathematical Reasoning via Formal Verification
Mistral leverages Lean 4 and synthetic data to move beyond natural language math, creating a model focused on formally verified proofs.
Industry
Google DeepMind and A24 Partner to Integrate AI into Filmmaking
A critical look at the partnership between Google DeepMind and A24, exploring whether it's artistic collaboration or a corporate data acquisition play.
Policy
The UK's Generational Tobacco Ban: A Failure of Policy and Logic
An analysis of why the UK's generational smoking ban is likely to fail by ignoring human behavior and fueling black markets.
Research
The Office Comprehension Bench: Moving Beyond Text-Based File Parsing
The Office Comprehension Bench (OCB) highlights the failure of text-flattening and the critical need for native format parsing in LLM systems.
Industry
Microsoft Launches AI Deployment Company to Bridge the Enterprise Gap
Microsoft invests $2.5 billion to launch a deployment arm, aiming to solve the 'last mile' problem of integrating AI into corporate environments.
Models
The Convergence of LLMs: Addressing the Problem of AI Groupthink
An exploration of how RLHF and synthetic data loops are creating homogenized AI outputs and the challenges of introducing divergent thinking.
Research
The Gap Between Formal Multilingual AI and Romanized Indic Code-Mixing
The Indi-RomCoM benchmark reveals that LLMs struggle with Romanized Indic-English code-mixing, highlighting a critical failure in current multilingual training pipelines.
Policy
Anthropic's Fable and Mythos Models: Global Release and Policy Implications
An analysis of the global release of Anthropic's Fable and Mythos models and the US government's shift in regulatory approach.
Industry
Anthropic's Claude Science and the Shift Toward Vertical AI Models
An analysis of Claude Science and whether specialized vertical models can solve the grounding problem in drug discovery and biotechnology.